Research On Streaming Media QoS Multicast Routing In Software-Defined Networking

With the rapid development of network scale and multimedia technology,streaming media services such as telemedicine,industrial Internet of Things,video conferencing have become popular on the Internet,and video traffic transmission is playing an important role in network.Moreover,with the rapid enrichment of network services and the consumption of resources,such as delay increase,network congestion and other problems,which reduce the QoS of multimedia services and traditional Internet network finally cannot content the needs of diversified multimedia services.Therefore,it is imperative to explore a new network architecture.Software-Defined Networking(SDN)as a new network architecture,realizes the decoupling and communication of data forwarding and control functions by using Open Flow protocol,provides a more fine-grained QoS guarantee,and improves the service quality of multimedia services.Therefore,this thesis studies the QoS multicast routing problem of streaming media in SDN architecture,so as to better guarantee the QoS requirements of multimedia services,and proposes the following two schemes:(1)Scalable QoS multicast routing for SVC streams in SDN.Scalable video coding is a kind of technology that can layer video coding in terms of time,quality and space.This thesis proposes a salable QoS multicast routing scheme for SVC streams.This scheme first presents the design of multimedia multicast system in SDN architecture,then proposes the time-efficient and scalable video layer selection algorithm to construct multicast tree for traffic routing,and further considers the dynamic process of client joining,video layer switching strategy and client leaving.Finally,the prototype system and the simulation experiment are deployed.The experimental results prove that the proposed scheme can content the QoS requirements and improve the scalability of the network.(2)QoS routing optimization scheme based on deep reinforcement learning in SDN.Deep reinforcement learning combines deep neural network with reinforcement learning to solve complex network routing problems.This thesis proposes a QoS routing optimization scheme based on deep reinforcement learning.Firstly,the multi-media multicast system is proposed based on the distributed control plane architecture.Then,the environment aware module in the controller is used to collect the underlying network state information,and the deep deterministic strategy gradient algorithm is deployed to the QoS module.The results demonstrate that the scheme can avoid network congestion and obtain approximate optimal routing strategy,which further verifies the effectiveness and feasibility of the scheme.